In the tech world, there are always some eye-catching developments, and this time it's OpenAI and Microsoft's AI heavyweights grabbing the spotlight. Word has it that Microsoft's AI Vice President Sebastien Bubeck is set to光荣离职, preparing to join OpenAI to continue pursuing the elusive dream of AGI (Artificial General Intelligence). Over here, the recently departed OpenAI Chief Technology Officer Mira Murati is also planning her own startup venture, with rumors that she intends to poach talent from OpenAI to establish her new company.
Bubeck, a prominent figure at Microsoft for a decade, previously served as an assistant professor at Princeton University. His achievements are numerous; most recently, he led the development of the Phi series of models, which have performed exceptionally well, even outperforming GPT-3.5 in various tests. The Phi-1 model made its debut with an impressive performance, truly stunning. Subsequently, the team continuously upgraded, releasing the Phi-1.5, Phi-2, and Phi-3 series, especially Phi-3.5, which not only streamlined parameters but also introduced the MoE architecture, surpassing GPT-4o in some benchmark tests.
Bubeck is full of anticipation for the future, expressing his hope to delve deeper into "how intelligence emerges in large language models (LLMs)" and plans to use this understanding to enhance model performance. Although he is leaving Microsoft, other collaborators in the team will continue to work on developing the Phi series.
Meanwhile, former CTO Murati is seeking her own entrepreneurial opportunities. She may start a new company with former OpenAI Vice President Barret Zoph or join other tech companies. According to insider information, some OpenAI researchers are interested in Murati's startup plans and may even follow her departure. This has caused some turbulence within the OpenAI team, with some researchers requesting transfers.
Additionally, Murati's departure is not alone; recently, other OpenAI employees have also left, including ChatGPT developer Luke Metz. Their exodus could pose greater challenges for OpenAI, especially in the increasingly fierce competition for high salaries.
Coincidentally, OpenAI's Vice President of Security Research, Lilian Weng, also posted a job opening on the same day, hoping to recruit new research scientists and engineers, despite external doubts about her capabilities. Overall, OpenAI is facing the dual pressure of talent loss, and the future situation looks bleak.
Paper address: https://arxiv.org/pdf/2404.14219